Farb et al. (2010) — Mindfulness training alters the neural expression of sadness
The empirical paper the wiki’s whole Farb dual-mode story was waiting for without quite having: the one where the protective sensory pole is interoceptive — the right insula — rather than the exteroceptive vision that Farb et al. (2011) measured and had to apologize for. It completes Week 11 The Absence of Emotion (the third and last file, after the 2011 and 2022 relapse papers), and it is chronologically the earliest of the three Farb data papers on the wiki, so it reads as the origin of a programme the wiki met out of order.
Design
Thirty-six participants enrolling in an 8-week Mindfulness-Based Stress Reduction programme (MBSR) at St. Joseph’s Health Centre, Toronto, were randomly assigned to a mindfulness-training (MT) group (n=20, scanned after completing MBSR) or a waitlisted control group (n=16, scanned before attending). During fMRI they watched sad and neutral film clips (film-based mood provocation; sad = “The Champ”, “Terms of Endearment”; neutral = gardening/woodworking) and rated sadness 1–5 after each clip. Emotional reactivity was BOLD(sad) − BOLD(neutral); a between-groups ANOVA on those difference maps isolated regions where reactivity was sensitive to training.
The design’s fault line is that it is not a within-subject pre/post: MT is scanned late, control early. Random assignment protects group equivalence at baseline (age, gender, BDI/BAI/SCL-90 all ns) but confounds “trained” with “scanned second.” Every reading below is hostage to that caveat, and the authors say so.
The core result: two networks, and training keeps the body one online
In controls, sadness provocation did two things at once (see the paper’s Figure 1):
| network | regions | reads as | |
|---|---|---|---|
| activated by sadness | cortical-midline + left language | vmPFC (BA11), dmPFC (BA9/32), posterior cingulate → precuneus (BA23/29/31), left DLPFC/Broca’s, Wernicke’s/STS | self-referential elaboration, rumination, “giving sorrow words” |
| deactivated by sadness | right viscerosomatic / experiential | right insula, subgenual ACC (BA25), somatosensory (BA3/4/5), right LPFC | body-sensation / interoceptive representation switching off |
So the untrained response to a sad film is to talk to yourself about it while the felt body drops out — the elaborative-vs-sensory trade-off Farb draws everywhere, here caught as a live activation/deactivation dyad in one contrast.
MT flattened both halves. The trained group showed less midline/language activation (reduced precuneus, mPFC, Broca’s, Wernicke’s) and less viscerosomatic deactivation — the experiential network’s signature training effect was recovery of the right insula (MT>control, peak Z=4.07) and right subgenual ACC/gyrus rectus. ROI analyses confirmed the deactivating regions were driven by generalized greater reactivity in controls (both larger neutral-film activation and larger sad-film deactivation vs. the resting baseline), which was absent in MT. The trained brain doesn’t recruit the ruminative network as hard and doesn’t shut the body-sensation network off.
Why this paper matters more to the wiki than its size suggests
Because of the correlation the wiki keeps asking 2011 and 2022 for and not getting: right insula reactivity was negatively correlated with depression (BDI r(34)=-.465, p<.005; right LPFC r=-.406). The participants who stayed least dysphoric under the sad challenge were the ones recruiting the interoceptive insula the most. This is the wiki’s first first-hand Farb datum in which the protective sensory pole is genuinely interoceptive — the insula, not the calcarine cortex — and it is tied to a clinical scale.
Set against it, a mirror finding: left Wernicke’s-area (STS) reactivity correlated positively with BDI (r=.542, p<.001), and the insula/LPFC (interoceptive) and Wernicke’s (language) ROIs were negatively correlated (r=-.501). More language-laden processing of the sadness went with more depression and less interoceptive engagement; MT lowered reactivity in both and softened the opposition. Read through the wiki’s framework, this is the perceptual-inference (stay with the sensation) vs. active-inference (elaborate/regulate it) contrast, with depression loading on the elaborative side and interoception on the protective side — the pattern 2011 would recover prospectively with an exteroceptive sensory pole.
The honest boundary (which way it cuts, this time)
The wiki’s standing complaint about the Farb relapse papers is that their protective “sensory” pole is exteroceptive, so the interoceptive reading is a generalization. This paper is the exception: the insula is interoception’s cortical home (insular-cortex), so here the sensory-pole-is-interoceptive claim is measured, not extrapolated. What this paper lacks instead is what those papers have — a prospective clinical outcome. The BDI correlation is cross-sectional and, worse, is contaminated by the between-groups design (MT participants both had lower BDI and were the trained/second-scanned group, so the insula-BDI link pools a within-group and a between-group source of variance). So the two limitations are complementary across the Farb trio: 2010 has the right region (insula) but a weak design (cross-sectional, confounded); 2011/2022 have strong designs (prospective, RCT) but the wrong region (vision, somatosensory). The interoception-first thesis is best supported by the trio together, and no single one of them nails it — a point worth keeping visible rather than smoothing.
Relation to Farb et al. (2007)
This is the emotion-provocation companion to Farb, Segal, Mayberg et al. (2007, SCAN), the narrative-vs-experiential self-reference study the wiki cites through norman-farb and simulation-map but does not hold in raw/. 2007 showed the two self-reference modes at rest / during a labeling task; 2010 shows the same right-lateralized viscerosomatic vs. midline-elaborative split emerging under emotional challenge and being reshaped by training. The paper leans on 2007 explicitly for the interpretation that MT shifts self-representation from midline-affective toward lateral-viscerosomatic circuitry.
Provenance
Norman Farb is this wiki’s author (see norman-farb); recorded as provenance, not weighted as authority, per the convention used for farb-2011-relapse-prediction and payne-2015-somatic-experiencing. Senior author Zindel Segal is the MBCT co-developer and clinical anchor. Co-author Helen Mayberg — whose subgenual-ACC / deep-brain-stimulation work on depression underlies this paper’s reading of the BA25 finding — is a prominent figure but enters only as a co-author here and is not given her own page, following the wiki’s co-author convention (as with Adam K. Anderson). The paper declares no conflicts of interest for the authors; Mayberg reports IP-licensing income from Advanced Neuromodulation Systems.